Unix pipes in multidimensional NMR data acquisition

David Naugler (dnaugler@sfu.ca)
Sun, 19 May 1996 20:34:43 -0700 (PDT)

Our Bruker AMX600 is set up to acquire data in only one way, with all
multidimensional data stored in a serial file. This imposes severe
restrictions when setting up an experiment. There must be a better way. I
can imagine a setup that can be described using Unix pipes. I would like
to know if anything like this has been implemented.

I believe it can be shown that digital oversampling can somewhat improve
the dynamic range, signal to noise and flatness of the baseline which is
important in nD NMR. Providing permanent storage for all that data imposes
an impossible limitation however.

Unix pipes are a nice way of processing large amounts of data without
requiring large amounts of temporary storage. It must be an important
concept because one NMR processing package is called NMRpipe.

While Bruker's method of storing large serial files works OK for 2D work,
it can be argued that the method is fundamentally unsound for nD work.
The correct way (possibly the DMX way) can be represented by pipes:

PP | TD | t3FT | PK | ST > file

for a 3D experiment, where, in the above:

PP = nD Pulse program producing stream of FID's, with digital
oversampling at ~ 100,000 complex samples per second.

TD = time domain processing e.g. digital filtering for removal of solvent
signal.

t3FT = Fourier Transform in t3.

PK = phase correction.

ST = extraction of strip transform and other data compression.

file = a smaller matrix (t2, t1) file of complex (compressed) spectra of
the region of interest.

Has anyone implemented a data acquisition scheme like this? Can this
scheme be implemented on a Bruker AMX600?

David Naugler
Institute of Molecular Biology and Biochemistry
Simon Fraser University